论文部分内容阅读
利用南开大学语音实验室开发的计算机语音工作站(MiniSpeechLab),对12个人的语音样本进行了长时平均功率谱(LTAS)分析。对于同一发音人不同内容、不同长度的语音文本、同一语音文本不同发音人以及男女发音人的 LTAS 进行了比较鉴别,提取了比较的均方差和最大能量差,对LTAS 在说话人识别中的作用进行了归纳总结。实验结果表明,话者语音 LTAS 抗干扰能力较强,体现个人特征稳定,能表现出说话人之间的个体差异,该参数可以用于说话人的语音鉴定。
Using the MiniSpeech Lab developed by Nankai Phonetic Lab, long-term average power spectrum (LTAS) analysis was performed on 12 people’s speech samples. For different contents of the same speaker, different length of the voice text, the same voice pronunciation of different speakers and male and female speaker LTAS are compared to identify the comparison of the mean square error and the maximum energy difference, the LTAS in the role of speaker recognition Conducted a summary. The experimental results show that LTAS has strong anti-jamming ability, stable personal characteristics and individual differences between speakers. This parameter can be used for speech recognition of the speaker.